Closed tdammers closed 5 years ago
Perhaps fixed by https://gitlab.com/gitlab-org/gitlab-ce/merge_requests/21728 ?
Looks very much like it. The only problem, then, is when trac users differ only by case, but are actually distinct users, with distinct mail addresses. The way things are now, we would silently merge them, and retain only the email address of whoever came first. I don't think this is what we want.
On Thu, Feb 21, 2019, 14:30 Matthew Pickering <notifications@github.com wrote:
Perhaps fixed by https://gitlab.com/gitlab-org/gitlab-ce/merge_requests/21728 ?
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/bgamari/trac-to-remarkup/issues/37#issuecomment-465999895, or mute the thread https://github.com/notifications/unsubscribe-auth/ABXjxxni6sApc7NIL572OZVEbl5HTO_nks5vPp92gaJpZM4bHQdG .
I believe this is sorted.
When creating new users, gitlab will refuse the request if a user with the given username or e-mail address already exists.
The logs reveal that occasionally, this happens even when just prior we queried for the exact same username and email address, and didn't find anything for either, which suggests that both the name and the address don't exist in gitlab yet. But still it fails.
Further evidence: the usernames in question do actually exist in the database, but casing differs (e.g.
samb
vs.SamB
).Hypothesis: gitlab, at least the API, is inconsistent wrt case sensitivity in usernames and passwords; it seems that querying is case sensitive, while the uniqueness check is not.
Unfortunately, I don't have a good solution for this right now.